Neurobiologically-Based Subtyping of Multi-Cohort Samples with MDD and PTSD Symptoms

摘要 创伤后应激症状重叠率高，症状重叠率高 精神障碍(PTSD)和严重抑郁障碍(MDD)让人质疑这两种疾病是否是截然不同的障碍。 这两种综合征的发病和病程都受到环境变量的强烈影响。我们假设 生活压力或逆境的连续体和独立的心理创伤连续体共同作用 影响创伤后应激障碍和MDD的发病(创伤后应激障碍至少需要一次创伤暴露)。我们的 首要目标是识别和比较MDD、PTSD、症状特征的共同特征 创伤后应激障碍和MDD，以及迄今未被识别的神经生物学定义的综合征。因此，我们计划 使用有监督的学习研究神经特征，并确定跨越障碍的生物类型(PTSD 和MDD)和无监督学习，这种方法可以更好地解释创伤、压力 生活事件和疾病特征比DSM障碍可能的更多。而不是对患者进行亚型 根据临床症状或DSM定义的诊断，我们的目标是识别不同的集群 来自静息状态功能磁共振成像的神经信号中断的神经生物学亚型。在目标1中，我们建议 使用监督学习来训练算法，以从可以分类的静止fMRI数据中检测神经特征 DSM诊断共患PTSD和MDD，仅PTSD，仅MDD，以及对照组(无精神障碍)。 分析将分别与经历过标准A级创伤或 有压力的生活事件，以及那些没有压力的人。在目标2中，我们计划在MDD和PTSD中使用监督学习 患者从与四种跨诊断相关的静息状态fMRI数据中识别神经特征 症状包括睡眠中断、易怒、注意力不集中和兴趣丧失。在目标3中，我们 建议应用无监督学习方法来识别与特定症状相关的新生物类型 或症状群。该算法将在(1)仅有PTSD、(2)仅有MDD的患者中使用rsfMRI特征 以及(2)PTSD、MDD和PTSD+MDD并存的患者，以确定潜在的跨诊断 跨越DSM边界的生物型。我们将调查诊断特异性和反式-DNA之间的关系 诊断生物型源于压力生活事件、创伤暴露的无监督学习， 接触时的发育阶段、精神并存、内科并存、疾病慢性病、 疾病严重程度、性别和年龄。将电路中断映射到的重叠和交叉模式 在设计在电路层面进行干预的治疗方法方面，精神病综合征是一个令人望而生畏的挑战。 开发一种基于神经生物学的病因学，映射到临床症状和综合征，代表着一种 翻译神经科学的重大进展。现代脑刺激技术的出现提供了一种 通过精准医疗策略在电路层面进行干预的可能性前所未有。

ABSTRACT Significant symptom overlap and high rates of co-occurrence between syndromes of posttraumatic stress disorder (PTSD) and major depressive disorder (MDD) call into question whether the two are distinct disorders. The onset and course of both syndromes are strongly influenced by environmental variables. We hypothesize that a continuum of life stress or adversity and an independent continuum of psychological trauma conspire to influence the onset of PTSD and MDD (where at least one trauma exposure is required for PTSD). Our overarching goal is to identify and compare neural signatures of MDD, PTSD, symptom features common to PTSD and MDD, and heretofore unrecognized neurobiologically-defined syndromes. Therefore, we plan to investigate neural signatures with supervised learning, and to identify biotypes that cut across disorders (PTSD and MDD) with unsupervised learning, an approach that can better explain contributions of trauma, stressful life events, and disease characteristics than possible with DSM-disorders. Rather than subtyping patients on the basis of clinical symptoms or DSM-defined diagnoses, our goal is to identify distinct clusters of neurobiological subtypes with disrupted neural signatures derived from resting-state fMRI. In Aim 1 we propose to train algorithms with supervised learning to detect neural signatures from resting fMRI data that can classify DSM diagnosis of comorbid PTSD and MDD, PTSD only, MDD only, and Controls (no psychiatric disorder). The analysis will be performed separately with MDD and Control groups who experienced criterion-A trauma or stressful life events, and those who did not. In Aim 2, we plan to use supervised learning in MDD and PTSD patients to identify neural signatures from resting-state fMRI data associated with four trans-diagnostic symptoms that include disrupted sleep, irritability, concentration difficulties, and loss of interest. In Aim 3, we propose to apply unsupervised learning methods to identify novel biotypes associated with specific symptoms or symptom clusters. The algorithms will employ rsfMRI features in patients with (1) PTSD only, (2) MDD only and (2) across PTSD, MDD, and comorbid PTSD+MDD patients in order to identify potential trans-diagnostic biotypes that cut across DSM boundaries. We will investigate associations of diagnosis-specific and trans- diagnostic biotypes derived from unsupervised learning with stressful life events, trauma exposure, developmental stage at time of exposure, psychiatric comorbidities, medical comorbidities, illness chronicity, illness severity, gender, and age. The overlapping and intersecting patterns that maps circuit disruption to psychiatric syndromes presents a daunting challenge in designing treatments that intervene at the circuit level. Developing a neurobiologically-based nosology that maps to clinical symptoms and syndromes represents a major advance in translational neuroscience. The advent of modern brain stimulation technology offers an unprecedented possibility of intervening at the circuit level with precision medicine strategies.